Definition

AI-Generated Virtual Environments refers to AI systems that generate complete, interactive virtual environments for consumption by humans, AI agents, and machines. This encompasses two distinct technical paradigms:

  1. World Models that understand dynamics and physics and generate interactive simulations
  2. 3D Asset/Environment Generators that produce exportable geometry, textures, and scene data compatible with game engines

The domain is characterized by:

  • Rapid technological evolution (2024-2025 saw the emergence of “Large World Models” as a distinct category)
  • Significant architectural diversity (video-based vs. geometry-based approaches)
  • Fundamental questions about what “generation” means when outputs range from real-time video streams to batch-produced mesh files

World Model Disambiguation

The term “world model” itself has two related but distinct meanings:

TypeDescriptionExamples
Generative World ModelsCreate external environmentsWorld Labs Marble, Google Genie, NVIDIA Cosmos
Internal Predictive World ModelsReasoning systems for AI agentsMeta V-JEPA

This report focuses on Generative World Models.


In Scope

The following are explicitly within the scope of this research:

  • AI systems that generate complete environments, levels, or worlds (batch and runtime)
  • Outputs intended for interactive consumption in engines and runtimes (Unreal, Unity, web viewers)
  • Geometry-native outputs (meshes, textures) and neural-native outputs (Gaussian splats)
  • Multi-user contexts where players and agents consume the same world state
  • Hybrid workflows combining procedural graphs with neural generation
  • Generation metadata, reproducibility, and validation profiles
  • Multiplayer determinism requirements for AI-generated content

Out of Scope

TopicReasonCovered By
Agent cognition and decision-makingDistinct domain with different standards needsMSF Autonomous Agents Use Case
Provenance and attribution standardsInfrastructure dependencyMSF Transparency and Provenance
Cross-platform asset format standardsGeneral interchange, not AI-specific3D Asset Interoperability WG
NeRF representationsNo portable format, not real-time viableAcademic research
AR-specific concerns (spatial anchoring)Distinct requirementsMSF AR Use Cases 18-21
Robotics-only training scenariosNot intended for metaverse consumptionIndustrial simulation standards

Adjacent Domains

DomainRelationshipOverlap Areas
Robotics/Physical AIShared technology, distinct purposeWorld models (Cosmos), training environments
Digital TwinsOverlapping if multi-consumerIndustrial metaverse, enterprise simulation
Gaming/Game EnginesCore overlapPrimary deployment target
VFX/Film ProductionAdjacent marketPre-visualization, USD workflows
Procedural Content GenerationEvolution fromHybrid AI+PCG workflows
3D Asset GenerationComponent/subsystemSingle-object generation within environments

Key Terminology

World Model

AI system that understands dynamics and generates interactive simulations

Also known as:
  • Large World Model (LWM)
  • Foundation World Model
3D Gaussian Splatting (3DGS)

Explicit particle representation using millions of Gaussian distributions for real-time rendering

Also known as:
  • Gaussian Splats
  • 3DGS
  • GS
Spatial Intelligence

Capacity for AI to understand and generate spatially consistent 3D representations

Also known as:
  • 3D Intelligence
  • Spatial AI
Physical AI

NVIDIA term for AI systems designed for physics-aware robotics and simulation

Also known as:
  • Embodied AI
  • Robotics AI
Inverse Graphics

3D asset generation approach that reconstructs 3D from 2D images

Also known as:
  • Reconstruction
  • Structure from Image
Next-Token Prediction

World model approach that predicts future video frames from current state

Also known as:
  • Video Prediction
  • Frame Generation
Neural Assets

3D content stored as neural network weights rather than traditional geometry

Also known as:
  • Neural Representations
  • Radiance Fields
Generation Recipe

Complete parameter set needed to reproduce AI-generated content (prompt, seed, model version)

Also known as:
  • Generation Metadata
  • Reproducibility Data